Deep learning for automated lesion detection in mammography has gained widespread attention due to its potential to reduce the time needed for radiologists to detect lesions accurately. Deep learning models are now being widely used in automated lesion detection in digital mammography. These models are able to classify suspicious regions that refer to symptoms of breast cancer better than classical image processing techniques. However, traditional models are limited by the lack of large public datasets that are representative of the variety of lesions that can exist in mammograms. Recent works have proposed deep learning models trained on large public datasets such as Digital Database for Screening Mammography (DDSM) and Digital Mammographic Image Archive (MI-AMI) that have achieved promising performance. These deep learning models use imaging and anatomical information for lesion detection, which results in better performance than classical approaches. Future research could extend the use of deep learning by exploring new vector-based approaches, integrating more data from mammography and clinic specialties, and performing holistic analyses to gain better insights..